1 intro

1.1 Purpose

  1. Preliminary cross-task RSA: see also https://osf.io/72khu/
  2. Assess impact of encoding direction on patterns: systematic bias? magnitude?
  3. Help figure out appropriate way(s) to analyze these data for this question.

1.2 Notes

Data/sample/preproc

  • UB55 sans 432332
  • surface + fMRIprep
  • schaefer 400-07

1st-Level GLM

  • 1trpk, runwise GLMs, shifted events

Target TRs:

  • Axcpt: 7, 8, 9
  • Cuedts: 9, 10
  • Stern: 11, 12
  • Stroop: 2, 3, 4

GLM contrasts:

  • Axcpt: \(\text{BY} - \text{BX}\)
  • Cuedts: \((\text{InConSwitch} + \text{InConRepeat} - \text{ConSwitch} - \text{ConRepeat})/2\)
  • Stern: \(\text{LL5RN} - \text{LL5NN}\)
  • Stroop: \((\text{PC50InCon} + \text{biasInCon} - \text{PC50Con} - \text{biasCon})/2\)

RSA

Similarity measure:

  • Standard, “vanilla” RSA was used. That is, not a “cross-validated” RSA (e.g., cv-Euclidean, cv-Mahalanobis; Wather et al. 2016).
  • Similarity measure: standard linear correlation
  • NOT cross-valdiated (e.g., not cv-Euclidean/Mahalanobis)
  • NOT prewhitened
  • however, conducted in BETWEEN-RUN and WITHIN-RUN manner.

Models/Factors:

  • Run/encoding direction: 1/AP, 2/PA
  • Task: Axcpt, …, Stroop
  • Condition: high, low
  • fit jointly via multiple regression
  • response variable atanh transformed

Plotting and statistical details

  • all p-value corrections were performed whole-cortex, with FDR.
  • inferential tests used wilcoxon-signed rank

2 quick look: all correlations

2.1 univariate

2.2 bivariate / contrasts of interest

2.2.1 Between-run models

2.2.2 Within-run models

3 Parcel-wise stats

3.1 bivariate / contrasts of interest

3.1.1 Between-run models

3.1.2 Within-run models

3.2 brains

3.2.1 wn

3.2.1.1 hi0

3.2.1.2 hilo

3.2.1.3 hitask

3.2.1.4 lo0

3.2.1.5 lotask

3.2.1.6 task0

3.2.2 bt

3.2.2.1 hi0

3.2.2.2 hilo

3.2.2.3 hitask

3.2.2.4 lo0

3.2.2.5 lotask

3.2.2.6 task0